Genetic Polymorphisms in the Catechol Estrogen Metabolism Pathway and Breast Cancer Risk

A vast accumulation of data has shown that estrogen-related exposures play a role in breast cancer etiology (1, 2). Conceivably, this could be due in part to the mutagenic effects of the intermediate metabolites of estrogen within the catechol estrogen (CE) pathway. There is evidence that genotoxicity may operate through the formation of reactive estrogen metabolites, namely CE semiquinones and quinones that damage DNA via the formation of superoxide radicals and depurinating DNA adducts, although it is unclear what proportion of estradiol is converted into CEs (3-6).

It is plausible that the genotoxic potential of estrogen varies across individuals and may be influenced by genetic variation within the CE metabolism pathway. Specifically, CYP1B1 plays a role in the conversion of 17β-estradiol (E₂) into 4-OHE₂ CE (3, 5, 7-9). Several single nucleotide polymorphisms (SNP) within CYP1B1 have been shown to have functional effects on the catalytic properties of the CYP1B1 enzyme (9, 10), and variant alleles of CYP1B1*2 (Ala¹¹⁹Ser) and CYP1B1*3 (Leu⁴³²Val) have been observed to be associated with breast cancer risk in some (11, 12), but not all, previous studies (13-15). Catechol-O-methyltransferase (COMT) is involved in methylating (and thereby inactivating) CEs (5, 16). The COMT Met¹⁵⁸ allele has been hypothesized to produce an enzyme with reduced functionality, although previous epidemiologic studies have not shown an association with breast cancer (9, 14). If CEs are inactivated, they can be easily oxidized into CE semiquinones and then into CE quinones (5). Members of the glutathione S-transferase (GST) family are thought to play a role in the conjugation of CE quinones, with GSTP1 being the predominant GST enzyme found in the breast (16, 17). The GSTP1 Val¹⁰⁵ allele has been shown to confer different catalytic enzyme activity (16). However, the epidemiologic findings regarding breast cancer risk have been mixed (14, 18-21). Null mutations have been characterized for both GSTM1 and GSTT1. For GSTM1, one pooled analysis reported a marginally significant, modestly increased risk of breast cancer associated with the null variant (14), although no increased risk was observed in a meta-analysis (22). For the GSTT1 null variant, no pooled analyses have observed an association with breast cancer risk (14).

The amount of CE accumulating in the breast tissue from daily exposure to estrogen supplied by estrogen-progestin therapy (EPT) or estrogen-alone therapy (ET) may conceivably vary among postmenopausal women and may be affected by the variation in genes within the CE metabolism pathway. Several studies have investigated interactions between genes in this pathway, as well as gene-EPT interactions, but no specific gene-gene combination has been observed to be associated with breast cancer in more than one study (19-21, 23-25). We investigated whether variation within five genes in the CE metabolism pathway was associated with an increased risk of breast cancer and whether EPT and/or ET modified the effect of genotype on breast cancer risk.

The study participants were recruited for a population-based case-control study, the Puget Sound Area Breast Cancer Evaluation (PACE) study of invasive breast cancer among postmenopausal women of 65 to 79 y of age in western Washington State. The methods have been described in detail previously (26) and thus are summarized only briefly here. Eligible case participants were (a) women 65 to 79 y of age when diagnosed with primary, invasive breast cancer between April 1, 1997, and May 31, 1999; (b) residents of the three county (King, Pierce, and Snohomish counties) Seattle-Tacoma metropolitan area at diagnosis; (c) women with no previous history of in situ or invasive breast cancer; and (d) women with the ability to communicate in English. All cases were ascertained through the Cancer Surveillance System, the population-based tumor registry serving the Seattle-Puget Sound region of western Washington State and a participant in the Surveillance, Epidemiology, and End Results Program of the National Cancer Institute. The PACE study interviewed 975 cases (80.6%) of those identified through the Cancer Surveillance System.

Eligible control participants for the PACE Study were identified through the Center for Medicare and Medicaid Services' list of Social Security recipients from the general population of the same three county Metropolitan Seattle area from which the cases were drawn. The Center for Medicare and Medicaid Services Social Security list includes all individuals 65 y and older residing in the United States who are eligible for Medicare benefits, including women whose income precludes eligibility to receive funds and women who elect not to receive cash benefits. The controls for this study were randomly selected from the Center for Medicare and Medicaid Services rolls and were frequency-matched to the expected distribution of cases by 5-y age group. Of the 1,365 controls selected, 1,007 women (73.8%) were interviewed.

Data collection for cases and controls was done in an identical manner. For each participant, upon the attainment of informed consent, a structured, in-person interview was conducted by a trained study interviewer on established and suspected breast cancer risk factors, including demographic characteristics, reproductive history, menstrual history, hormonal contraception history, medical history, certain medications, weight and height history, lifestyle factors, family history of cancer, and hormonal therapy use. The questions on hormonal therapy were extensive and solicited information on lifetime use, including drug name, start and stop dates of use, separate and combined uses of estrogen and progestin, strength, and monthly pattern of pill use. All interview questions were limited to events occurring before each participant's diagnosis date (reference date for controls). Within the 5-y age groups on which controls were frequency-matched to cases, reference dates for controls were assigned in a distribution matching the cases' expected diagnosis dates.

Blood samples were provided by 891 (91.4%) of the interviewed cases and 878 (87.1%) of the interviewed controls. This study was approved by the institutional review board of the Fred Hutchinson Cancer Research Center.

DNA was extracted from buffy coats using manual phenol chloroform method. Genotyping was done at the Functional Genomics Laboratory of the Center for Ecogenetics and Environmental Health at the University of Washington.

The 5′-nuclease TaqMan Detection System–based assays were developed to discriminate the following alleles: CYP1B1*2 (Ala¹¹⁹Ser), CYP1B1*3 (Leu⁴³²Val), COMT Val¹⁵⁸Met, and GSTP1 Ile¹⁰⁵Val (Supplementary Table S1A). Primers and dual-labeled allele-specific probes were designed through Assays-by-DesignSM Service-SNP Genotyping by Applied Biosystems, Inc. Each TaqMan minor groove binder probe consisted of an oligonucleotide labeled both with a particular 5′ reporter dye and a 3′ nonfluorescent quencher. Amplification was done by initial denaturation at 95°C for 10 min, followed by 40 cycles of amplification at 92°C for 15 s and annealing at 60°C for 1 min. End-point analysis was done on ABI 7900 Sequence Detection System (Applied Biosystems, Inc.) to determine the genotypes.

The presence or absence of the GST mu (GSTM1) and theta (GSPT1) genes was determined using a multiplex PCR (27). Briefly, two sets of primers were used to amplify a 215-bp segment of the GSTM1 gene and a 480-bp segment of the GSTT1 gene. Primers were also included to amplify a 268-bp segment of the β-globin gene, which was used as a positive control for the PCR. The products were separated by electrophoresis with ethidium bromide–stained 2.5% agarose gel (ISC BioExpress) and genotyped by visual inspection using UV illumination.

Positive controls consisting of DNA aliquots representing wild-type/wild-type, wild-type/mutant, mutant/mutant genotypes (characterized by DNA-sequencing), and a negative control (no DNA) were included in each assay. In addition, randomly selected samples (10%) were identified for replicate testing and were integrated randomly into the genotyping plates. Laboratory staff members were blinded to case-control status and to replicate status.

Among the 10% of the samples for which genotypes were replicated for quality control assessment, the level of agreement between the samples was 100% for the COMT Val¹⁵⁸Met alleles, 100% for CYP1B1*2 Ala¹¹⁹Ser, 94.7% for CYP1B1*3 Leu⁴³²Val, 93.2% for GSTM1*0, 96.2% for GSTP1 Ile¹⁰⁵Val, and 96.4% for GSTT1*0. When discordance occurred, the genotype was coded as a missing variable. Additionally, the amount of missing data (including those removed for quality control purposes) for each genotype was 0.0% (COMT), 2.0% (CYP1B1*2), 0.6% (CYP1B1*3), 0.5% (GSTM1), 0.2% (GSTP1), and 0.2% (GSTT1).

The EPT variable was categorized as never use, short-term use (<60 mo of use), and long-term use (=60 mo of use). The ET variable was similarly categorized as never use, short-term use, and long-term use. In addition, analysis examining ET use excluded any women who had ever used EPT due to the association observed between EPT use and breast cancer risk in this data set.

Concordance between replicate samples included for quality control purposes was determined for each SNP. Testing for deviation from Hardy-Weinberg equilibrium was done within population controls using a χ² goodness of fit test. Mantel-Haenszel χ² test was used for tests of associations in bivariate analyses.

Odds ratios (OR) and the corresponding 95% confidence intervals (95% CI) were calculated using unconditional logistic regression. ORs were adjusted for age and year of diagnosis because these were frequency-matched variables. The Wald's test statistic was used to test trends in models containing variables with multiple levels. We also used the Wald's test statistic to determine if the trend in ORs associated with various genotypes differed by hormone therapy use.

In the multigene analysis, we considered only SNPs statistically significantly associated with breast cancer in the single gene analyses. Similarly, only when a single-gene or multigene model was statistically significantly associated with breast cancer risk did we investigate the potential for a gene-hormone therapy interaction. The exception to this was the investigation of interactions that have been previously reported in the literature.

Interaction terms were investigated using the likelihood ratio test. We estimated CYP1B1 haplotype frequencies and ORs and 95% CIs associated with breast cancer. Haplotype estimation was done using Phase v.2.1. The CYP1B1 haplotype with the highest frequency (G at A119S and G at L432V) served as the reference category.

We accounted for multiple testing in our analyses by using the false-positive reporting probability (FPRP) and preset the FPRP-level criterion at 20% (based on the number of studies previously investigating these genes and our adequate sample size). We evaluated the FPRP using prior probabilities ranging from 0.1 to 0.01 (28).

Demographic and hormone-related characteristics of breast cancer cases and controls are presented in Table 1. Similar to findings previously reported in the overall PACE study (26), cases had a higher body mass index and were more likely to be long-term users of EPT than controls.

In the single-gene analyses, women homozygous for the CYP1B1*2 Ser allele (T/T) had a 1.69-fold increased risk of breast cancer (95% CI, 1.17-2.46) and the heterozygotes (G/T) were at no increased risk (OR, 0.99; 95% CI, 0.81-1.20) compared with homozygous wild-types (Table 2). Women homozygous for the GSTP1 Val allele (G/G) had a 0.73-fold increased risk of breast cancer (95% CI, 0.54-0.99) and the heterozygotes (A/G) had no altered risk (OR, 1.04; 95% CI, 0.85-1.27) compared with the homozygous wild-type. The COMT Met¹⁵⁸ allele (rs4680), CYP1B1 Val⁴³² allele (rs1056836), GSTM1 null allele, and GSTT1 null allele were not observed to be associated with risk of breast cancer.

We investigated the possibility of reporting a false-positive association between breast cancer and the SNPs in our study by calculating a FPRP for prior probabilities ranging from 0.1, 0.05, to 0.01. We observed FPRPs equal to 0.11, 0.21, and 0.58, at these prior probability levels, respectively, for CYP1B1*2, and FPRPs equal to 0.06, 0.11, and 0.39, respectively, for GSTP1, indicating that the FPRP was within the 20% criterion set a priori for both of these SNPs at a prior probability of 0.1 but not at 0.01.

There was no indication that the associations seen for the homozygous variant allele of CYP1B1*2 differed according to the use of EPT or ET (Table 3). For GSTP1, the reduced risk of breast cancer among women homozygous for the Val allele was limited to never users of ET (P for interaction = 0.004). Although we had data on duration and recency of use, our sample size was too limited to investigate gene-disease associations stratified by EPT duration and recency.

The two genes found to have associations with breast cancer, CYP1B1*2 and GSTP1, were assessed for a possible gene-gene interaction. However, we observed no evidence of this (P = 0.36; Table 4). We also observed no elevation in risk associated with the CYP1B1*2-CYP1B1*3 haplotypes (compared with GG: OR, 1.0; 95% CI, 0.8-1.1, for GC; OR, 1.0; 95% CI, 0.8-1.3, for TG; OR, 1.1; 95% CI, 0.9-1.3, for TC; data not shown).

Additionally, we investigated possible interactions between genes previously reported in the literature (19-21, 23-25) but did not detect any among the following combined gene variables: GSTT1-GSTM1, COMT-GSTP1, COMT-GSTT1, COMT-GSTM1, COMT-GSTM1-GSTT1, and GSTP1-GSTM1-GSTT1 (Supplementary Table S1).

We also investigated a possible EPT interaction with combined genes as reported by Mitrunen et al. and did not observe evidence that the risk of breast cancer associated with combined genes varied according to EPT use within the COMT-GSTP1 combined variable (P = 0.19; Supplementary Table S2). With the COMT-GSTT1 and COMT-GSTM1 combined variables, although we observed a greater risk in short-term users of EPT than was predicted by their separate associations (P = 0.004 and 0.001, respectively), the pattern of risk when stratified by EPT was not what we would have expected a priori (comparing 2 with 0 high-risk genotypes in COMT-GSTT1: OR, 0.8; 95% CI, 0.5-1.4, for never users; OR, 24.4; 95% CI, 3.0-200.2, for EPT use <30 months; OR, 0.8; 95% CI, 0.3-2.4, for EPT use of 30+ months; in COMT-GSTM1: OR, 0.7; 95% CI, 0.4-1.1, for never users; OR, 8.5; 95% CI, 1.4-52.4, for EPT use <30 months; and OR, 1.3; 95% CI, 0.7-2.3, for EPT use of 30+ months), admittedly with a constrained sample size for this analysis.

Our results were unchanged when limited to White women for all of the analyses presented above.

Our study observed that variants in the CYP1B1 and GSTP1 genes were associated with breast cancer risk. In interpreting these results, we must first consider the potential limitations of our study. Laboratory errors resulting in nondifferential misclassification of genotypes could bias ORs toward the null. However, quality control analyses indicated high correlations between replicate samples. Also, allele distributions were consistent with Hardy-Weinberg equilibrium and there was a minimal amount of missing genotype data.

Another potential limitation was the omission of a portion of the interviewed cases (8.6%) and controls (12.9%) who were otherwise eligible for the parent study but did not donate a blood specimen. Comparison of the women who donated blood with all participants interviewed is reassuring in that there were no discernible differences in risk-factor distributions and risk estimates, nor did the cases who donated blood vary from the entire case series in terms of stage and other disease features. Another possible source of selection bias, however, could arise from the 19.4% of cases and 26.2% of controls who were identified by the Cancer Surveillance System but did not participate in the parent study. For analyses that consider interview data, the ability of women to recall past exposures is a potential concern. Specific tools designed to assist recall were used in this study, including a lifetime calendar and colored photographs of medications. As a check on the quality of recall in PACE participants, we previously compared self-reported data on several categories of medications with pharmacy records and found overall agreement to be quite good (29).

A large number of individual studies have reported on associations between genes in this pathway and breast cancer risk (11-13, 15, 18, 20-25, 30-36). However, no genome-wide association studies have reported findings for regions encompassing these genes (37, 38). In addition to several large individual studies, a meta-analysis of CYP1B1*2 found no association between breast cancer risk and the Ala/Ser genotype [OR, 1.1 (95% CI, 0.9-1.2)], or the Ser/Ser genotype (OR, 1.0; 95% CI, 0.8-1.2; refs. 9, 11, 13, 15, 31, 33). Findings from haplotype analyses have not been any more compelling. Among three large studies investigating the risk of breast cancer associated with haplotypes in the CYP1B1 gene, one observed a 1.5-fold increased risk comparing women homozygous for G,T,C at positions 48, 119, and 432, respectively, with those with the most common haplotype (95% CI, 1.0-2.1; P = 0.03). The two other studies reported no alteration in breast cancer risk in any CYP1B1 haplotype (15, 31, 33). The generally negative findings from previous studies argue that the 1.7-fold increased risk of breast cancer that we observed among women with the CYP1B1*2 Ser allele should be interpreted with caution.

For GSTP1, a pooled analysis of 301 cases and 397 controls reported that the G/G genotype (Val homozygotes) was associated with an increased risk of breast cancer (OR, 1.86; 95% CI, 1.05-3.3; ref. 14), but two larger studies have observed either a decreased risk (OR, 0.6; 95% CI, 0.3-1.0) or no association among the Val homozygotes (OR, 0.9; 95% CI, 0.5-1.4; refs. 19, 20). However, among postmenopausal women, these studies have observed borderline decreased risks associated with Val homozygotes compared with wild-type homozygotes (OR, 0.5; 95% CI, 0.2-1.1 and OR, 0.7; 95% CI, 0.4-1.1; refs. 19, 20). In the two studies that are stratified by use of hormone therapy, one reported a greater reduction in breast cancer risk observed among ever users of ET (OR, 0.2; 95% CI, 0.04-0.8) than never users (OR, 0.6; 95% CI, 0.2-2.0; ref. 20). However, another study observed no differential association with Val homozygosity according to use of EPT (OR, 0.8; 95% CI, 0.5-1.2, for never users; OR, 0.8; 95% CI, 0.4-1.5, for ever users; ref. 20).

Our study, the largest-to-date study examining the association between GSTP1 and breast cancer risk, observed women with the GSTP1 Val/Val genotype generally to be at a reduced risk of breast cancer, and is in broad agreement with the findings of other epidemiologic studies reported for postmenopausal women. We evaluated the probability that this finding represented a false-positive association and calculated the likelihood of that to be 11% at a prior probability of 0.05 (at lower prior probabilities, the false-reporting probability is higher). Thus, if the prior probability of an association between the GSTP1 Val/Val allele is at least 5% (given the a priori biological rationale for investigating SNPs within this gene, it is reasonable to assume the prior probability is 5% or greater), the probability of this finding being a false-positive association is relatively small.

Additionally, although it is plausible that the reduction in breast cancer risk associated with the GSTP1 Val/Val genotype is most pronounced in never users of ET, these findings would need to be replicated before any inferences can be made. Adding complexity to the issue are the conflicting findings demonstrating varying levels of catalytic activity between the variant protein and the wild-type protein. Some showed a higher V_max for the Val¹⁰⁵ isoform, whereas others showed a lower catalytic efficiency for the Val¹⁰⁵ isoform compared with wild-type (39-42).

Our null findings with COMT, GSTM1, and GSTT1 and breast cancer are in broad agreement with previous reports from pooled and meta-analyses (9, 14, 22). Also, if account is taken of the effect modification by age in the HuGE review of CYP1B1*3 (for Val/Leu compared with Leu/Leu genotypes in Caucasians women >60 years of age: OR, 1.1; 95% CI, 0.8-1.4; 51-54 years of age: OR, 2.5; 95% CI, 1.1-5.7; 55-59 years of age: OR, 1.9; 95% CI, 1.1-3.6; ref. 43), then our findings in this study of women ages 65 years and older are consistent with previous works as well.

Previous studies have also investigated possible gene-gene combinations within this set of genes as there is strong biological rationale for considering the joint effects of genes within this shared pathway, with several studies (18, 20, 23-25, 44) but not all (21), reporting associations between high-risk genotypes from combined genes in this pathway and breast cancer risk. However, with the exception of two studies reporting marginally significant increased risks of breast cancer associated with combined GSTP1, GSTM1, and GSTT1, no specific gene-gene combination has been observed in more than one study (18, 20, 23-25, 44). We investigated but did not detect any gene-gene interactions within the CE pathway.

A previous study reported on a potential interaction between combined gene variables and EPT in breast cancer (24). Specifically, among EPT users of >30 months, Mitrunen et al. observed the risk of breast cancer to be heightened in relation to COMT-GSTM1 (OR, 9.1; 95% CI, 1.8-45.0), COMT-GSTT1 (OR, 8.4; 95% CI, 1.4-49.0), and COMT-GSTP1 (OR, 7.0; 95% CI, 1.2-40.6) comparing 2 with 0 high-risk genotypes. An increased risk was not similarly observed among EPT users of <30 months (COMT-GSTM1: OR, 0.7; 95% CI, 0.2-3.5, comparing 2 with 0 high-risk genotypes; COMT-GSTT1: OR, 1.7; 95% CI, 0.3-10.5; and COMT-GSTP1: OR, 2.3; 95% CI, 0.4-14.1). Although we observed a greater risk associated with two high-risk genotypes for short-term users of EPT for both the COMT-GSTT1 combined variable and the COMT-GSTM1 combined variable, these patterns of risk were not what we had hypothesized a priori (nor were the patterns in agreement with those reported by Mitrunen et al.), and thus, we cannot draw any conclusions based on these findings.

Our study observed a 27% reduction in breast cancer risk associated with the GSTP1 Val/Val genotype. However, we are the first study to report a statistically significant decreased risk among postmenopausal women; clearly, replication of our results is needed before any firm conclusion can be drawn. Based on this study and the existing epidemiologic literature, there is little evidence that any SNPs in the CE metabolism pathway, with the exception of GSTP1, have a main effect on breast cancer risk and there is little evidence supporting interactions between any of these genes and hormone therapy use.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.